/* drain the send queue */
static void rxe_qp_drain(struct rxe_qp *qp)
{
if (qp->sq.queue) {
if (qp->req.state != QP_STATE_DRAINED) {
qp->req.state = QP_STATE_DRAIN;
if (qp_type(qp) == IB_QPT_RC)
rxe_run_task(&qp->comp.task, 1);
else
__rxe_do_task(&qp->comp.task);
rxe_run_task(&qp->req.task, 1);
}
}
}
/* move the qp to the error state */
void rxe_qp_error(struct rxe_qp *qp)
{
qp->req.state = QP_STATE_ERROR;
qp->resp.state = QP_STATE_ERROR;
/* drain work and packet queues */
rxe_run_task(&qp->resp.task, 1);
if (qp_type(qp) == IB_QPT_RC)
rxe_run_task(&qp->comp.task, 1);
else
__rxe_do_task(&qp->comp.task);
rxe_run_task(&qp->req.task, 1);
}
void rnr_nak_timer(unsigned long data)
{
struct rxe_qp *qp = (struct rxe_qp *)data;
pr_debug("qp#%d rnr nak timer fired\n", qp_num(qp));
rxe_run_task(&qp->req.task, 1);
}
void rnr_nak_timer(struct timer_list *t)
{
struct rxe_qp *qp = from_timer(qp, t, rnr_nak_timer);
pr_debug("qp#%d rnr nak timer fired\n", qp_num(qp));
rxe_run_task(&qp->req.task, 1);
}
static int rxe_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
struct ib_send_wr **bad_wr)
{
int err = 0;
struct rxe_qp *qp = to_rqp(ibqp);
unsigned int mask;
unsigned int length = 0;
int i;
int must_sched;
if (unlikely(!qp->valid)) {
*bad_wr = wr;
return -EINVAL;
}
if (unlikely(qp->req.state < QP_STATE_READY)) {
*bad_wr = wr;
return -EINVAL;
}
while (wr) {
mask = wr_opcode_mask(wr->opcode, qp);
if (unlikely(!mask)) {
err = -EINVAL;
*bad_wr = wr;
break;
}
if (unlikely((wr->send_flags & IB_SEND_INLINE) &&
!(mask & WR_INLINE_MASK))) {
err = -EINVAL;
*bad_wr = wr;
break;
}
length = 0;
for (i = 0; i < wr->num_sge; i++)
length += wr->sg_list[i].length;
err = post_one_send(qp, wr, mask, length);
if (err) {
*bad_wr = wr;
break;
}
wr = wr->next;
}
/*
* Must sched in case of GSI QP because ib_send_mad() hold irq lock,
* and the requester call ip_local_out_sk() that takes spin_lock_bh.
*/
must_sched = (qp_type(qp) == IB_QPT_GSI) ||
(queue_count(qp->sq.queue) > 1);
rxe_run_task(&qp->req.task, must_sched);
return err;
}
void retransmit_timer(unsigned long data)
{
struct rxe_qp *qp = (struct rxe_qp *)data;
if (qp->valid) {
qp->comp.timeout = 1;
rxe_run_task(&qp->comp.task, 1);
}
}
static int rxe_post_send_kernel(struct rxe_qp *qp, const struct ib_send_wr *wr,
const struct ib_send_wr **bad_wr)
{
int err = 0;
unsigned int mask;
unsigned int length = 0;
int i;
while (wr) {
mask = wr_opcode_mask(wr->opcode, qp);
if (unlikely(!mask)) {
err = -EINVAL;
*bad_wr = wr;
break;
}
if (unlikely((wr->send_flags & IB_SEND_INLINE) &&
!(mask & WR_INLINE_MASK))) {
err = -EINVAL;
*bad_wr = wr;
break;
}
length = 0;
for (i = 0; i < wr->num_sge; i++)
length += wr->sg_list[i].length;
err = post_one_send(qp, wr, mask, length);
if (err) {
*bad_wr = wr;
break;
}
wr = wr->next;
}
rxe_run_task(&qp->req.task, 1);
if (unlikely(qp->req.state == QP_STATE_ERROR))
rxe_run_task(&qp->comp.task, 1);
return err;
}
static void rxe_skb_tx_dtor(struct sk_buff *skb)
{
struct sock *sk = skb->sk;
struct rxe_qp *qp = sk->sk_user_data;
int skb_out = atomic_dec_return(&qp->skb_out);
if (unlikely(qp->need_req_skb &&
skb_out < RXE_INFLIGHT_SKBS_PER_QP_LOW))
rxe_run_task(&qp->req.task, 1);
rxe_drop_ref(qp);
}
/* rxe_recv calls here to add a request packet to the input queue */
void rxe_resp_queue_pkt(struct rxe_qp *qp, struct sk_buff *skb)
{
int must_sched;
struct rxe_pkt_info *pkt = SKB_TO_PKT(skb);
skb_queue_tail(&qp->req_pkts, skb);
must_sched = (pkt->opcode == IB_OPCODE_RC_RDMA_READ_REQUEST) ||
(skb_queue_len(&qp->req_pkts) > 1);
rxe_run_task(&qp->resp.task, must_sched);
}
void rxe_comp_queue_pkt(struct rxe_dev *rxe, struct rxe_qp *qp,
struct sk_buff *skb)
{
int must_sched;
skb_queue_tail(&qp->resp_pkts, skb);
must_sched = skb_queue_len(&qp->resp_pkts) > 1;
if (must_sched != 0)
rxe_counter_inc(rxe, RXE_CNT_COMPLETER_SCHED);
rxe_run_task(&qp->comp.task, must_sched);
}
static inline enum comp_state complete_ack(struct rxe_qp *qp,
struct rxe_pkt_info *pkt,
struct rxe_send_wqe *wqe)
{
unsigned long flags;
if (wqe->has_rd_atomic) {
wqe->has_rd_atomic = 0;
atomic_inc(&qp->req.rd_atomic);
if (qp->req.need_rd_atomic) {
qp->comp.timeout_retry = 0;
qp->req.need_rd_atomic = 0;
rxe_run_task(&qp->req.task, 1);
}
}
if (unlikely(qp->req.state == QP_STATE_DRAIN)) {
/* state_lock used by requester & completer */
spin_lock_irqsave(&qp->state_lock, flags);
if ((qp->req.state == QP_STATE_DRAIN) &&
(qp->comp.psn == qp->req.psn)) {
qp->req.state = QP_STATE_DRAINED;
spin_unlock_irqrestore(&qp->state_lock, flags);
if (qp->ibqp.event_handler) {
struct ib_event ev;
ev.device = qp->ibqp.device;
ev.element.qp = &qp->ibqp;
ev.event = IB_EVENT_SQ_DRAINED;
qp->ibqp.event_handler(&ev,
qp->ibqp.qp_context);
}
} else {
spin_unlock_irqrestore(&qp->state_lock, flags);
}
}
do_complete(qp, wqe);
if (psn_compare(pkt->psn, qp->comp.psn) >= 0)
return COMPST_UPDATE_COMP;
else
return COMPST_DONE;
}
static inline enum comp_state complete_wqe(struct rxe_qp *qp,
struct rxe_pkt_info *pkt,
struct rxe_send_wqe *wqe)
{
qp->comp.opcode = -1;
if (pkt) {
if (psn_compare(pkt->psn, qp->comp.psn) >= 0)
qp->comp.psn = (pkt->psn + 1) & BTH_PSN_MASK;
if (qp->req.wait_psn) {
qp->req.wait_psn = 0;
rxe_run_task(&qp->req.task, 1);
}
}
do_complete(qp, wqe);
return COMPST_GET_WQE;
}
static int rxe_post_recv(struct ib_qp *ibqp, const struct ib_recv_wr *wr,
const struct ib_recv_wr **bad_wr)
{
int err = 0;
struct rxe_qp *qp = to_rqp(ibqp);
struct rxe_rq *rq = &qp->rq;
unsigned long flags;
if (unlikely((qp_state(qp) < IB_QPS_INIT) || !qp->valid)) {
*bad_wr = wr;
err = -EINVAL;
goto err1;
}
if (unlikely(qp->srq)) {
*bad_wr = wr;
err = -EINVAL;
goto err1;
}
spin_lock_irqsave(&rq->producer_lock, flags);
while (wr) {
err = post_one_recv(rq, wr);
if (unlikely(err)) {
*bad_wr = wr;
break;
}
wr = wr->next;
}
spin_unlock_irqrestore(&rq->producer_lock, flags);
if (qp->resp.state == QP_STATE_ERROR)
rxe_run_task(&qp->resp.task, 1);
err1:
return err;
}
static int rxe_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
const struct ib_send_wr **bad_wr)
{
struct rxe_qp *qp = to_rqp(ibqp);
if (unlikely(!qp->valid)) {
*bad_wr = wr;
return -EINVAL;
}
if (unlikely(qp->req.state < QP_STATE_READY)) {
*bad_wr = wr;
return -EINVAL;
}
if (qp->is_user) {
/* Utilize process context to do protocol processing */
rxe_run_task(&qp->req.task, 0);
return 0;
} else
return rxe_post_send_kernel(qp, wr, bad_wr);
}
/*
* IBA Spec. Section 10.7.3.1 SIGNALED COMPLETIONS
* ---------8<---------8<-------------
* ...Note that if a completion error occurs, a Work Completion
* will always be generated, even if the signaling
* indicator requests an Unsignaled Completion.
* ---------8<---------8<-------------
*/
static void do_complete(struct rxe_qp *qp, struct rxe_send_wqe *wqe)
{
struct rxe_cqe cqe;
if ((qp->sq_sig_type == IB_SIGNAL_ALL_WR) ||
(wqe->wr.send_flags & IB_SEND_SIGNALED) ||
wqe->status != IB_WC_SUCCESS) {
make_send_cqe(qp, wqe, &cqe);
advance_consumer(qp->sq.queue);
rxe_cq_post(qp->scq, &cqe, 0);
} else {
advance_consumer(qp->sq.queue);
}
/*
* we completed something so let req run again
* if it is trying to fence
*/
if (qp->req.wait_fence) {
qp->req.wait_fence = 0;
rxe_run_task(&qp->req.task, 1);
}
}
int rxe_completer(void *arg)
{
struct rxe_qp *qp = (struct rxe_qp *)arg;
struct rxe_dev *rxe = to_rdev(qp->ibqp.device);
struct rxe_send_wqe *wqe = wqe;
struct sk_buff *skb = NULL;
struct rxe_pkt_info *pkt = NULL;
enum comp_state state;
rxe_add_ref(qp);
if (!qp->valid || qp->req.state == QP_STATE_ERROR ||
qp->req.state == QP_STATE_RESET) {
rxe_drain_resp_pkts(qp, qp->valid &&
qp->req.state == QP_STATE_ERROR);
goto exit;
}
if (qp->comp.timeout) {
qp->comp.timeout_retry = 1;
qp->comp.timeout = 0;
} else {
qp->comp.timeout_retry = 0;
}
if (qp->req.need_retry)
goto exit;
state = COMPST_GET_ACK;
while (1) {
pr_debug("qp#%d state = %s\n", qp_num(qp),
comp_state_name[state]);
switch (state) {
case COMPST_GET_ACK:
skb = skb_dequeue(&qp->resp_pkts);
if (skb) {
pkt = SKB_TO_PKT(skb);
qp->comp.timeout_retry = 0;
}
state = COMPST_GET_WQE;
break;
case COMPST_GET_WQE:
state = get_wqe(qp, pkt, &wqe);
break;
case COMPST_CHECK_PSN:
state = check_psn(qp, pkt, wqe);
break;
case COMPST_CHECK_ACK:
state = check_ack(qp, pkt, wqe);
break;
case COMPST_READ:
state = do_read(qp, pkt, wqe);
break;
case COMPST_ATOMIC:
state = do_atomic(qp, pkt, wqe);
break;
case COMPST_WRITE_SEND:
if (wqe->state == wqe_state_pending &&
wqe->last_psn == pkt->psn)
state = COMPST_COMP_ACK;
else
state = COMPST_UPDATE_COMP;
break;
case COMPST_COMP_ACK:
state = complete_ack(qp, pkt, wqe);
break;
case COMPST_COMP_WQE:
state = complete_wqe(qp, pkt, wqe);
break;
case COMPST_UPDATE_COMP:
if (pkt->mask & RXE_END_MASK)
qp->comp.opcode = -1;
else
qp->comp.opcode = pkt->opcode;
if (psn_compare(pkt->psn, qp->comp.psn) >= 0)
qp->comp.psn = (pkt->psn + 1) & BTH_PSN_MASK;
if (qp->req.wait_psn) {
qp->req.wait_psn = 0;
rxe_run_task(&qp->req.task, 1);
}
state = COMPST_DONE;
break;
case COMPST_DONE:
if (pkt) {
rxe_drop_ref(pkt->qp);
kfree_skb(skb);
skb = NULL;
}
goto done;
case COMPST_EXIT:
if (qp->comp.timeout_retry && wqe) {
state = COMPST_ERROR_RETRY;
break;
}
/* re reset the timeout counter if
* (1) QP is type RC
* (2) the QP is alive
* (3) there is a packet sent by the requester that
* might be acked (we still might get spurious
* timeouts but try to keep them as few as possible)
* (4) the timeout parameter is set
*/
if ((qp_type(qp) == IB_QPT_RC) &&
(qp->req.state == QP_STATE_READY) &&
(psn_compare(qp->req.psn, qp->comp.psn) > 0) &&
qp->qp_timeout_jiffies)
mod_timer(&qp->retrans_timer,
jiffies + qp->qp_timeout_jiffies);
WARN_ON_ONCE(skb);
goto exit;
case COMPST_ERROR_RETRY:
/* we come here if the retry timer fired and we did
* not receive a response packet. try to retry the send
* queue if that makes sense and the limits have not
* been exceeded. remember that some timeouts are
* spurious since we do not reset the timer but kick
* it down the road or let it expire
*/
/* there is nothing to retry in this case */
if (!wqe || (wqe->state == wqe_state_posted)) {
WARN_ON_ONCE(skb);
goto exit;
}
if (qp->comp.retry_cnt > 0) {
if (qp->comp.retry_cnt != 7)
qp->comp.retry_cnt--;
/* no point in retrying if we have already
* seen the last ack that the requester could
* have caused
*/
if (psn_compare(qp->req.psn,
qp->comp.psn) > 0) {
/* tell the requester to retry the
* send queue next time around
*/
rxe_counter_inc(rxe,
RXE_CNT_COMP_RETRY);
qp->req.need_retry = 1;
rxe_run_task(&qp->req.task, 1);
}
if (pkt) {
rxe_drop_ref(pkt->qp);
kfree_skb(skb);
skb = NULL;
}
WARN_ON_ONCE(skb);
goto exit;
} else {
rxe_counter_inc(rxe, RXE_CNT_RETRY_EXCEEDED);
wqe->status = IB_WC_RETRY_EXC_ERR;
state = COMPST_ERROR;
}
break;
case COMPST_RNR_RETRY:
if (qp->comp.rnr_retry > 0) {
if (qp->comp.rnr_retry != 7)
qp->comp.rnr_retry--;
qp->req.need_retry = 1;
pr_debug("qp#%d set rnr nak timer\n",
qp_num(qp));
mod_timer(&qp->rnr_nak_timer,
jiffies + rnrnak_jiffies(aeth_syn(pkt)
& ~AETH_TYPE_MASK));
rxe_drop_ref(pkt->qp);
kfree_skb(skb);
skb = NULL;
goto exit;
} else {
rxe_counter_inc(rxe,
RXE_CNT_RNR_RETRY_EXCEEDED);
wqe->status = IB_WC_RNR_RETRY_EXC_ERR;
state = COMPST_ERROR;
}
break;
case COMPST_ERROR:
WARN_ON_ONCE(wqe->status == IB_WC_SUCCESS);
do_complete(qp, wqe);
rxe_qp_error(qp);
if (pkt) {
rxe_drop_ref(pkt->qp);
kfree_skb(skb);
skb = NULL;
}
WARN_ON_ONCE(skb);
goto exit;
}
}
exit:
/* we come here if we are done with processing and want the task to
* exit from the loop calling us
*/
WARN_ON_ONCE(skb);
rxe_drop_ref(qp);
return -EAGAIN;
done:
/* we come here if we have processed a packet we want the task to call
* us again to see if there is anything else to do
*/
WARN_ON_ONCE(skb);
rxe_drop_ref(qp);
return 0;
}
static inline enum comp_state check_ack(struct rxe_qp *qp,
struct rxe_pkt_info *pkt,
struct rxe_send_wqe *wqe)
{
unsigned int mask = pkt->mask;
u8 syn;
struct rxe_dev *rxe = to_rdev(qp->ibqp.device);
/* Check the sequence only */
switch (qp->comp.opcode) {
case -1:
/* Will catch all *_ONLY cases. */
if (!(mask & RXE_START_MASK))
return COMPST_ERROR;
break;
case IB_OPCODE_RC_RDMA_READ_RESPONSE_FIRST:
case IB_OPCODE_RC_RDMA_READ_RESPONSE_MIDDLE:
if (pkt->opcode != IB_OPCODE_RC_RDMA_READ_RESPONSE_MIDDLE &&
pkt->opcode != IB_OPCODE_RC_RDMA_READ_RESPONSE_LAST) {
return COMPST_ERROR;
}
break;
default:
WARN_ON_ONCE(1);
}
/* Check operation validity. */
switch (pkt->opcode) {
case IB_OPCODE_RC_RDMA_READ_RESPONSE_FIRST:
case IB_OPCODE_RC_RDMA_READ_RESPONSE_LAST:
case IB_OPCODE_RC_RDMA_READ_RESPONSE_ONLY:
syn = aeth_syn(pkt);
if ((syn & AETH_TYPE_MASK) != AETH_ACK)
return COMPST_ERROR;
/* Fall through (IB_OPCODE_RC_RDMA_READ_RESPONSE_MIDDLE
* doesn't have an AETH)
*/
case IB_OPCODE_RC_RDMA_READ_RESPONSE_MIDDLE:
if (wqe->wr.opcode != IB_WR_RDMA_READ &&
wqe->wr.opcode != IB_WR_RDMA_READ_WITH_INV) {
return COMPST_ERROR;
}
reset_retry_counters(qp);
return COMPST_READ;
case IB_OPCODE_RC_ATOMIC_ACKNOWLEDGE:
syn = aeth_syn(pkt);
if ((syn & AETH_TYPE_MASK) != AETH_ACK)
return COMPST_ERROR;
if (wqe->wr.opcode != IB_WR_ATOMIC_CMP_AND_SWP &&
wqe->wr.opcode != IB_WR_ATOMIC_FETCH_AND_ADD)
return COMPST_ERROR;
reset_retry_counters(qp);
return COMPST_ATOMIC;
case IB_OPCODE_RC_ACKNOWLEDGE:
syn = aeth_syn(pkt);
switch (syn & AETH_TYPE_MASK) {
case AETH_ACK:
reset_retry_counters(qp);
return COMPST_WRITE_SEND;
case AETH_RNR_NAK:
rxe_counter_inc(rxe, RXE_CNT_RCV_RNR);
return COMPST_RNR_RETRY;
case AETH_NAK:
switch (syn) {
case AETH_NAK_PSN_SEQ_ERROR:
/* a nak implicitly acks all packets with psns
* before
*/
if (psn_compare(pkt->psn, qp->comp.psn) > 0) {
rxe_counter_inc(rxe,
RXE_CNT_RCV_SEQ_ERR);
qp->comp.psn = pkt->psn;
if (qp->req.wait_psn) {
qp->req.wait_psn = 0;
rxe_run_task(&qp->req.task, 1);
}
}
return COMPST_ERROR_RETRY;
case AETH_NAK_INVALID_REQ:
wqe->status = IB_WC_REM_INV_REQ_ERR;
return COMPST_ERROR;
case AETH_NAK_REM_ACC_ERR:
wqe->status = IB_WC_REM_ACCESS_ERR;
return COMPST_ERROR;
case AETH_NAK_REM_OP_ERR:
wqe->status = IB_WC_REM_OP_ERR;
return COMPST_ERROR;
default:
pr_warn("unexpected nak %x\n", syn);
wqe->status = IB_WC_REM_OP_ERR;
return COMPST_ERROR;
}
default:
return COMPST_ERROR;
}
break;
default:
pr_warn("unexpected opcode\n");
}
return COMPST_ERROR;
}
